Plsql Programming Guidelines within a RAC Environment

Question

Hi Guys,
Are there any specific guidelines you would recommend for programming within a RAC environment, in addition to the usual best practices one would adopt for writing Plsql programs? ie guidelines that would take advantage of the performance boost from the high availability feature of the RAC environment, as well as the processing power?

Answer 1

There are realistically no guidelines that are specific to RAC (other than making sure if you're referencing V$ tables in code that you're using the GV$ version if you want data from all the nodes).
On the other hand, there are plenty of guidelines that are nice in single-node environments that potentially become necessities in a RAC environment. In particular, all the guidelines that reduce contention are useful in a single-node environment but can become critically important in a RAC environment. Anything that causes hot blocks in a single-node system can cause massive scaling problems in a RAC environment.
- If you are using permanent tables rather than global temporary tables to store temporary data, once load crosses a critical point, that can cause every session in the database to be queued up waiting for some other session to finish with the handful of blocks in this table to get sent over the interconnect from a different node in the cluster. That will bring your system to its knees.
- If you are inserting monotonically increasing keys into an index (i.e. a sequence generated primary key), you want as many of those primary key indexes to be reverse key indexes as possible (where "possible" takes into account the downside of no longer being able to do range scans on those indexes) to minimize index hot blocks and thus contention for those blocks over the interconnect.
- etc.
Justin